A1073
Title: Quantifying uncertainty in electricity prices forecasting: Models and methods
Authors: Alessandro Giovannelli - University of L'Aquila (Italy) [presenting]
Tommaso Proietti - University of Roma Tor Vergata (Italy)
Andrea Cerasa - European Commission - Joint Research Centre (Italy)
Fany Nan - (Italy)
Abstract: The focus is on short-term electricity price forecasting. In particular, the objective is twofold: on the one hand, the performance of forecasting methods is documented with different assumptions (i.e., reduced forms, structural decompositions, nonlinearity) and degrees of mean reversion; secondly, it aims at exploring a procedure for interval prediction, concentrating on a new method, conformal prediction (CP), which is an effective procedure for distribution-free predictive inference in regression. The empirical application focuses on the prediction of the time series of the single national price of the Italian electricity spot market in the short run, i.e., for forecast horizons that are not larger than 14 days ahead. Regarding the point forecast, findings suggest that the best-performing models are robust autoregressive predictors and unobserved component models featuring local trends and seasonality, whereas nonlinear specifications do not show a comparative advantage. With respect to the construction of prediction intervals, results suggest that CP produces fairly accurate and reliable prediction intervals, especially when the prediction interval is the result of a combination of a large set of methods used in forecasting.
